THE ORIGIN OF IDEALITY FACTORS N > 2 OF SHUNTS AND SURFACES

IN THE DARK I-V CURVES OF SI SOLAR CELLS

O. Breitenstein, P. Altermatt*, K. Ramspeck*, and A. Schenk**

Max-Planck-Institut für Mikrostrukturphysik, Weinberg 2, D-06120 Halle, Germany, phone: +49-345-5582740, fax: +49-
345-5511223, mail: breiten@mpi-halle.mpg.de
*University of Hannover, Inst. Solid-State Physics, Dep. Solar Energy, Appelstr. 2, D-30167 Hannover, Germany; and
Institute for Solar Energy Research Hameln (ISFH), Am Ohrberg 1, D-31860 Emmerthal, phone: +49-511-7625505, fax:
+49-511-76219760 mail: pietro@nano.uni-hannover.de
**Integrated Systems Laboratory, ETH Zurich, Gloriastr. 35, 8092 Zurich, Switzerland, and Synopsys Switzerland LLC,
Affolternstr. 52, 8050 Zurich, Switzerland, phone: +41 44 632 66 89, mail: schenk@iis.ee.ethz.ch

ABSTRACT: So far, a general model for the explanation of non-linear shunts and edge currents, often showing
ideality factors n > 2, has been missing. Non-linear shunts like scratches and edge currents are the major source of the
recombination current of industrial crystalline silicon solar cells. Moreover, the reverse current of such cells behaves
always linear or even superlinear instead of saturation-type, as expected. We propose that the edge recombination
current and other non-linear shunts can be described as regions that are highly disturbed, even amorphous, crossing
the pn-junction. The recombination within such regions needs model descriptions that go beyond the SRH
approximation, because the density of defects is so high that carriers recombine via more than one defect (so-called
multi-level recombination). Two variants of coupled defect level recombination, both leading to idealiy factors larger
than 2, are realistically modelled by Sentaurus simulations. Under reverse bias, the high local density of states in the
defective material enables hopping conduction, which explains the linear reverse characteristics.
Keywords: Fundamentals, Silicon, Simulation, Shunts.

1 INTRODUCTION anywhere in the cell area that are responsible for current
with n > 2 [6]. We propose that already the elementary
The Shockley-Read-Hall (SRH) recombination recombination process at the extended defects may be
theory, which assumes recombination via isolated point saturable, which leads to the observed high ideality
defect levels, predicts an ideality factor of the factors. For explaining this saturation, models beyond the
recombination current of n = 2 or less (see e.g. McIntosh SRH approximation have to be considered, which
et al. [1]), and saturation-type reverse characteristics. becomes probable if recombination occurs via extended
Industrial crystalline solar cells usually deviate from this defects [7]. Results of the simulation of two special cases
prediction. Their recombination current is much higher of defect pair recombination, both leading to high ideality
than expected for the given lifetime and often shows an factors, will be introduced.
ideality factor larger than 2 at relative low voltages.
Moreover, their reverse characteristic is linear or even
superlinear. Especially the large ideality factors have 2 EXPERIMENTAL RESULTS
attracted attention already very early [2] and are still
under discussion. Previous explanations of a high ideality 2.1 Artificial shunts, dark forward characteristics
factor were based on trap-assisted tunneling or field-
enhanced recombination [3] via isolated point defect
levels. However, realistic simulations have shown that
the electric field in the depletion region is too low to
enable effective single-level trap-assisted tunneling. Only
if two-level recombination implying at least one shallow
level is considered, which becomes probable only for
high defect concentrations, trap-assisted tunneling
becomes effective and indeed may lead to large ideality
factors [4]. Another explanation of high ideality factors
was based on the emitter series resistance between the
contacts and the recombination site [5]. We will show in
the following that this series resistance is not sufficient to
explain large ideality factors at the observed low
voltages.
A key for the understanding of large ideality factors Fig. 1: AFM image of a 25x25 µm sized region of the
is the fact that such recombination currents are flowing scratch with 27 g load. The scratch is about 1 µm deep.
not homogeneously in the cell, but always at local sites,
e.g. in extended defects. For example, it had been Planar, monocrystalline silicon solar cells with the
realized by many authors that, in the absence of other edge of the pn-junction lying well-passivated below an
defects, the major source of currents with n > 2 is the oxide layer ("PERL"-type cells), behave like "ideal" cells,
edge, where the pn-junction crosses the surface. By lock- i.e. they don't show a recombination current with an
in thermography investigations it had been proven that, ideality factor larger than 2. By cleavage [7, 8] or
besides the edges, there are local non-linear shunts
scratching, typical non-linear shunts can be produced in a
controlled manner. Here we have made 1 mm long
scratches at the front surface of several 1 cm2 sized
"ideal" cells with well-defined loads from 6 to 27 g by
using a Vickers indentor at room temperature. In Fig. 1,
an AFM image is shown of a scratch performed with 27 g
load. It is apparent that, even for this scratch with the
highest load, the material still was deformed plastically.
In Fig. 2 the dark I-V characteristic and in Fig. 3 the bias-
dependent ideality factor of an "ideal" (virgin) cell are
shown, together with characteristics of 3 equivalent cells,
which had been diamond-scratched over 1 mm length
using three different loads. The virgin cell shows an
ideality factor below 1.5 over the whole bias range. The
curve for the 27 g load.
2.2 Artificial shunts, reverse characteristics
Fig. 4 shows the room temperature reverse
characteristics of the cells of Figs. 2 and 3 in linear scale.
Both the virgin and the 6g scratched cell show a
negligible reverse current. We see that the reverse
conductivity strongly correlates with the magnitude of the
recombination current under forward bias: the higher the
recombination current, the higher is the reverse
conductivity. The characteristics are nearly linear up to 5
V.
0.10

reverse current [mA]

scratch with 6 g load clearly increased the forward 0.08
current and the ideality factor, but the ideality factor did load 27g
not exceed n = 2. The scratches with 9g and 27 g load 0.06 load 9g
generated a further increase both of the recombination virgin, load 6g
current and of the ideality factor, exceeding n = 2 already 0.04
for biases below 0.1 V. Since our scratches of 1 mm
length are all lying in the middle between two grid lines 0.02
having a distance of 1 mm in a 100 Ω/□ emitter, the
emitter series resistance to the scratch Rs is well known to 0.00
be 25 Ω. So the characteristics can be Rs-corrected, 2 0 3 1 4 5
which is shown for the 27 g load as dotted lines in Figs. 2 reverse bias [V]
and 3. The Rs-correction slightly reduces the ideality Fig. 4: Reverse characteristics of cells of Fig. 1
factor for biases above 0.3 V, but the general behavior 100 50 0 -50 -100 T[°C] -150
remains. This proves that here the series resistance is not
1E-3
reverse current [A]

the primary reason for a high ideality factor.

0.01
load 27g
Rs-corrected load 27g, -1 V
load 9g 1E-4
1E-3
load 6g
1E-4 virgin 1E-5
forward current [A]

1E-5
1E-6
1E-6

1E-7 1E-7
1E-8
0.22 0.24 0.26 0.28 0.30
1/4 -1/4
1E-9 1/T [K ]
0.0 0.1 0.2 0.3 0.4 0.5 0.6
Fig. 5: T-depencence of the reverse current of the cell
bias [V] scratched with 27 g load, measured at -1 V reverse bias
Fig. 2: Measured forward I-V characteristics of a virgin
"ideal" cell and 3 equivalent cells, which are diamond The temperature dependence of the I-V
scratched with different loads. Dotted: Rs-corrected curve characteristics may point to the conduction mechanism.
for the 27 g load. In Fig. 5 the T-dependence of the reverse current of the
Rs-corrected cell with the 27 g scratch measured at -1V is shown. This
5.5
load 27g dependence is displayed logarithmically over 1/T1/4. It is
5.0 load 9g found that, at least in the range between -50°C and
4.5 load 6g +100°C, this plot produces a straight line. It is well
virgin known that such a dependence is expected for variable
4.0
ideality factor

range hopping conduction in a constant density of states

3.5 near the Fermi level according to Mott's theory [9]. This
3.0 type of conductivity has been observed regularly for
2.5 highly disturbed semiconductors and insulators, like
nano- and microcrystalline silicon layers [10].
2.0
1.5
1.0 3 MODELLING
0.0 0.1 0.2 0.3 0.4 0.5 0.6
bias [V] A general theory for recombination via coupled
Fig. 3: Bias-dependent ideality factors of the defect levels is not available yet. However, we present the
characteristics shown in Fig. 2. Dotted: Rs-corrected simulation using two special cases of recombination via
two levels: Schenk's Coupled defect level model [4] and
the Deep donor-acceptor-pair (DAP) recombination
proposed recently [7]. In both cases recombination paths
according to Fig. 6 are assumed, where the usual SRH
recombination paths via two neighboring defect levels are
extended by an inter-defect transfer path described by the
transfer coefficient c12 and a thermal emission rate e12
[4].
EC
cn2 cn1 en1 en2
E1
c12 e12
E2
cp2 cp1
ep1 ep2
EV
Fig. 6: Recombination scheme including inter-level
recombination and emission
3.1 Simulations with the Coupled defect level model
A basic assumption of the coupled defect level
model is that the inter-level transfer coefficient c12 is so
large that it is not rate-limiting. If at least one of the two
energy levels is shallow [4], its wave function is
sufficiently extended in space so that trap-assisted
tunneling becomes the dominant mechanism for its free
carrier capture process, which is assumed to be rate-
limiting [11]. This capture mechanism does not require
large electric fields and is strongly field-dependent,
leading to a saturation of this recombination path with
increasing forward bias, corresponding to decreasing
field strength. Fig. 7 shows Sentaurus [12] simulations
using this model to compute the dark I-V characteristics
together with the bias-dependent ideality factor for
various defects in the pn-junction, having various
energies of the shallow trap level Et and various defect
lifetimes τ [7].
Fig. 7: Simulated forward I-V characteristics and ideality
factors of a PERL cell with a defect modelled by the
coupled defect level model [4]. 0: Et = 50 meV, τ = 1e-4
s, 1: Et = 90 meV, τ = 8e-7 s, 2: Et = 145 meV, τ = 6e-8
s, 3: Et = 190 meV, τ = 4e-8 s [7].
3.2 Deep donor-acceptor-pair recombination
This model assumes that capture and emission of free
carriers in a single level follows the usual SRH theory,
but that the captured carriers can be transferred to a
neighboring defect. For creating a saturation effect, we
have to assume a distinct asymmetry between the electron
and hole capture coefficients, so that the recombination
occurs via a donor-acceptor-pair (DAP). Until now, DAP
recombination has only been observed in luminescence
of shallow levels. In our case also deep donors and
acceptors are considered. Their wave function is much
more localized than that of shallow levels. However,
deep defects that are charged have sufficiently extended
wave functions such that their local concentration may be
high enough in extended defects that inter-level transfer
becomes probable. Under forward bias, in the middle of
the depletion region, for each isolated level the capture of
carriers with the lower capture coefficient would be rate-
limiting. Here the inter-level recombination may provide
an alternative recombination path bridging this
bottleneck, which leads to an increased recombination
rate compared to single levels. However, this inter-level
recombination path is saturable, because the density of
recombination partners is limited. This is the basic reason
for the saturation behaviour in the Deep DAP
recombination model. In Fig. 8 results of our Sentaurus
simulations are shown. The donor and acceptor energies
were set to midgap, and the smaller of the two capture
coefficients was set to zero for both levels. We vary the
transfer coefficient c12 [cm-3s-1]. For small c12 a maximum
of the ideality factor appears at low voltage, and with
increasing c12 it shifts to higher voltages. The second
smaller maxima around 0.5 V are due to the series
resistance to the defect. The experimental results are not
perfectly modelled in this simple model. Simulations
with variable energy of the levels have shifted the
maximum of the ideality factor to even higher voltages.
For further refining this model, an ensemble of defect
pairs with various energies and transfer coefficients has
to be assumed.
0.01 5.0
1E-3 e12=2e19 4.5
1E-4 e12=1e18
4.0
Current [A/cm ]
2
e12=1e17
1E-5
ideality factor
e12=1e16 3.5
1E-6
virgin 3.0
1E-7
2.5
1E-8
1E-9 2.0
1E-10 1.5
1E-11 1.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6
bias [V]
Fig. 8: Simulated forward I-V characteristics and ideality
factors (dashed) of a cell with a defect modelled by the
Deep DAP recombination model. Both levels are
assumed to lie at midgap.
4 DISCUSSION AND CONCLUSIONS
We are presenting a unified explanation for all
aspects of the non-ideal behavior of the dark I-V
characteristics of most industrial silicon solar cells, which
are a high magnitude and ideality factor of the
recombination current and an essentially ohmic reverse
characteristic. These non-ideal current contributions are
all due to extended defects like the cell's edge or other
defects in the cell area like scratches crossing the pn-
junction. If these extended defects have a low local
density of defect states, they behave according to the
SRH recombination theory of isolated point defects, and
the I-V characteristics have n < 2. However, if the
extended defects have a high local density of defect
states, multi-level recombination becomes probable,
which may show an intrinsic saturation behavior as
demonstrated here in two special cases. Moreover, for a
high density of defect states, hopping conduction in the
defect volume may govern the reverse conductivity of the
devices. Our experimental results on scratches with
different loads, which are typical examples of extended
defects with controlled densities of states, strongly
support this hypothesis.
In a general theory for multi-level recombination at
extended defects, a basic question is how these
recombination levels can be described. Here we propose
that they should be modelled in analogy to the defect
levels in amorphous silicon. Indeed, usual edge opening
procedures like laser cutting, dicing, and reactive ion
etching may generate an amorphous layer at the surface.
Even cleaving causes plastic deformation at the cleave-
tip, and also scratching implies plastic deformation in
micro-regions (see Fig. 1), which converts crystalline
silicon into highly disturbed (up to amorphous) material.
Non-hydrogenated amorphous silicon is characterized by
a very high defect density in the order of 1020 cm-3. This
high defect density makes both multi-level recombination
(as the dominant recombination mechanism under
forward bias) and hopping conduction (as the dominant
conduction mechanism under reverse bias) very probable.
Note that for such a high defect density the Fermi level is
pinned, hence free carrier conduction is improbable in
this material. Our measured T-dependence of the reverse
conductivity of the scratched solar cells strongly supports
the hypothesis that this conductivity is due to amorphous
or at least highly disturbed silicon material bridging the
pn-junction.
Until now, only special cases of multi-level
recombination have been modelled, and well-known
DAP recombination features like Coulomb interaction
and the influence of the pair distribution function of
neighbouring defects have not been considered.
Therefore, our modellings do not exactly reproduce the
experimental results. So a general theory for describing
the electronic influence of highly disturbed regions
crossing a pn-junction, which describes both forward and
the reverse characteristics, is still awaiting completion.
DC 1996, pp. 573-576.
[4] A. Schenk, U. Krumbein, "Coupled defect-level
recombination: Theory and application to anomalous
diode characteristics", J. Appl. Phys. 78 (1995), pp.
3185-3192.
[5] K.R. McIntosh, C.B. Honsberg, "The influence of
edge recombination on a solar cell's I-V curve", Proc.
16th PVSEC, Glasgow 2000, pp. 1651-1654.
[6] O. Breitenstein, M. Langenkamp, J.P. Rakotoniaina,
J. Zettner, "The imaging of shunts in solar cells by
infrared lock-in thermography", Proc. 17th Eur. PVSEC,
Munich 2001, pp. 1499-1502.
[7] O. Breitenstein, P. Altermatt, K. Pamspeck, M.A:
Green, J. Zhao, A. Schenk, "Interpretation of the
commonly observed I-V characteristics of c-Si cells
having ideality factors larger than two", 4th WCPEC,
Hawaii 2006, proceedings in print.
[8] S.W. Glunz et al., "High-efficiency silicon solar cells
for low-illumination application, Proc. 29th IEEE PVSC,
New Orleans 2002, pp. 450-453.
[9] N.F. Mott, Metal-Insulator Transitions, Taylor &
Francis, London 1990
[10] Seung Yeop Myong, Koeng Su Lim, "Universal
single-phonon variable range hopping conduction for
inorganic semiconducting polycrystalline films", Appl.
Phys. Lett. 88 (2006), 222110
[11] A. Schenk, "A model for the field and temperature
dependance of Shockely-Read-Hall lifetimes in silicon",
Solid-State Electronics 35 (1992), 1585
[12] SENTAURUS, Synopsys Inc. Mountain View,
CA, www.synopsys.com/products/tcad/tcad.html

The authors are grateful to S. Glunz (ISE Freiburg) for

providing the "ideal" solar cells used here, to H. Leipner
(Univ. Halle) for performing the diamond scratches, to
M. Alexe (MPI Halle) for performing AFM investigation
of the scratches, and to R. Carius (FZ Jülich) for valuable
discussions about the physical properties of non-
hydrogenated amorphous silicon.

REFERENCES

[1] K.R. McIntosh, P.P. Altermatt, G. Heiser,

"Depletion-region recombination in silicon solar cells:
when does mDR = 2 ?'', Proc.16th EC Solar Energy
Conference, Glasgow, UK, pp. 251-254, 2000.
[2] H.J. Queisser, "Forward characteristics and
efficiencies of silicon solar cells", Solid-State-Electronics
5 (1962) 1-10.
[3] A. Kaminski, J.J. Marchand, H. El Omari, A.
Laugier, Q.N. Le, D. Sarti, "Conduction processes in
silicon solar cells", Proc. 25th IEEE PVSC, Washington